onset-detection

Removal test: onset-detection operates by computing a time-varying energy or spectral feature (envelope) and locating peaks above threshold. Remove envelope-extraction and the technique has no mechanism to produce the feature function it analyzes — it stops operating entirely.

Onset detection is a signal processing technique designed for the sake of signal-processing: it identifies temporal events in audio signals, a core analytical operation used across both general signal processing and music applications.

Onset detection uses spectral flux (the frame-to-frame change in spectral energy) as its primary onset function. Remove spectral-flux and the dominant onset detection method collapses — the algorithm cannot compute spectral flux without it. The removal test passes at object level.

Onset detection is designed and maintained for the sake of music information retrieval: identifying temporal events in audio is a core analytical operation used across MIR tasks (beat tracking, genre classification, transcription). Its purpose is to further MIR's operation as a specific domain application, not just general signal processing.

Removal test: beat-tracking operates by first detecting onsets and then analysing their temporal regularity to find beat positions. Remove onset-detection and the technique loses its primary input — the onset sequence — and has no mechanism to produce it.

Short-time energy is designed to serve onset detection — its primary use case in speech processing and MIR is detecting the onset of sounds and events. The removal test: remove onset detection and short-time energy still exists, but the designed purpose relationship holds. Law 8d: servant (short-time-energy)→master (onset-detection).

Spectral flux is built and maintained for the sake of onset detection — its design purpose is to measure frame-to-frame spectral change to locate note onsets. The servant points at the master purpose, Law 8d.